tained in the first experiment were applied as forecast samples, then the data from second experiment were regarded as test
samples.
3. RESULTS AND ANALYSIS
Spectral reflectance of water and
Microcystic aeruginosa
with different concentration were obtained by control experiments,
under the condition of water disturbance and stationary. Spectral reflectance of bands whose wavelength located between
350nm-1000nm was used. Meanwhile two different narrow band vegetation indices and corresponding broad band vegeta-
tion indices were calculated Table 1. Narrow band vegetation indices consisted of any two bands in the entire 651 narrow
bands while broad band vegetation indices were the combina- tions of Landsat ETM+ NIR and RED bands.
These specific spectral indices were:
Broad band ETM SR and NDVI indices;
Narrow SR and NDVI indices: any two bands in the entire 651 bands.
Vegetation indices Definition
SR Broad band
NIR SR
RED
Narrow band
j SR
ij i
NDVI Broad band
RED RED
NIR NDVI
NIR
Narrow band
j i
NDVI ij
j i
Table 1: Spectral vegetation indices used in this paper Where NIR = Landsat ETM+ B4 bands
RED = Landsat ETM+ B3 bands.
3.1 Correlation of single band reflectance and
Microcystic aeruginosa
chlorophyll-a
Figure 1 shows the correlation coefficient r between single band reflectance and
chlorophyll-a
under the condition of water disturbance and stationary.
The trend of r is very close to the
Microcystic aeruginosa
spec- tral reflectance. In the band intervals in which r decreased and
increased quickly,
Microcystic aeruginosa
spectral reflectance has the similar change. In this area, reflectance is very sensitive
to
chlorophyll-a
concentration. The location of 3 minimum reflectance is corresponding to 3 reflection troughs of
Microcys- tic aeruginosa
spectrum; the location of 3 maximum is corre- sponding to
Microcystic aeruginosa
spectral green range peak, secondary band peak, red edge respectively.
Figure 1 Correlation coefficient between
Microcystic aerugino- sa
spectral reflectance and
chlorophyll-a
concentration
3.2 Correlation of broad band vegetation indices and
Mi- crocystic aeruginosa
chlorophyll-a
Under the condition of water disturbance and stationary, corre- lation coefficient r0.9 of
Microcystic aeruginosa chloro-
phyll-a
with band 2, band 4 are both very high. At band 1 and band 3, correlation coefficient r0.5 is low under the condition
of water disturbance. On the contrary, r is more than 0.85 under the condition of water stationary Figure 2.
The independent variables include single band reflectance from 733nm to 794nm, ETM
’s band 4 reflectance whose central band is 835nm, the SR and NDVI consisted of ETM
’s band 3 4 and ETM
’s band 3 band 4 central bands. With these variables, linear prediction model of
chlorophyll-a
was constructed. The precision was evaluated by calculating determination coefficient
and RMSE Table 2.
prediction Test
Independent variable
condition function
R
2
RMSE 799nm
D y=8010.8x-112.55
0.9850 84.45
34.15 85.43
31.76 794nm
S y =2355.8x-13.65
0.9903 835nm
D y=10090x-116.83
0.9805 S
y=2369.2x-2.2073 0.9912
This contribution has been peer-reviewed. doi:10.5194isprsarchives-XLI-B7-91-2016
94
ETM_B4 D
y=9818x-116.6 0.9812
87.23 32.34
30.9
53.82
31.96
53.46 113.8
133.7 111.6
133 S
y=2373x-3.518 0.9912
ETM_SR D
y=341.4x-233.5 0.9928
S y=124.8x-98.87
0.9921 C_SR
D y=364.19x-241.6
0.9956
S y=127.98x-98.01
0.9921 ETM_NDVI
D y=1276.4x+125.7
0.8693 S
y=985.4x-68.04 0.7193
C_NDVI D
y=1293.5x+149.92 0.8776
S y=983.1x-56.508
0.7258
Table 2: linear model ETM_SR and ETM_NDVI were used to represent broad band
indices and C_SR and C_NDVI to narrow band indices. Water disturbance and stationary was abbreviated as D and S respec-
tively. The relation between
chlorophyll-a
real value Y and pre- dicted value y is:
y a Y b
4
Where a= gain b = bias
The RMSE is regarded as a scale used to compare each model ’s
predictive ability. By comparing RMSE, under the condition of water disturbance, the predictive precision of SR model is high-
er, even though the R
2
of single band model and SR model are both very high. Hence there is almost little discrepancy between
narrow SR model and broad band SR model. Under the condi- tion of water stationary, SR model
’s predictive precision is low- er than that single band model
’s, where RMSE of SR model is 1.7 times higher than that of single band model. The predictive
ability of ETM+ ’s band 4 model is close to which of 794nm
model.
3.3 Correlation of narrow band vegetation indices and
